tracing: Use kstrdup_const instead of private implementation
authorRasmus Villemoes <linux@rasmusvillemoes.dk>
Wed, 9 Sep 2015 21:24:01 +0000 (23:24 +0200)
committerSteven Rostedt <rostedt@goodmis.org>
Thu, 1 Oct 2015 13:05:17 +0000 (09:05 -0400)
commit79ac6ef521075d0f40805df77e8890c55f538fe4
treecf353ec399027ddd9f93a958004cc61777c1d9a9
parent37aea98b84c0ce2ac638510fefeed9f8f920bd34
tracing: Use kstrdup_const instead of private implementation

The kernel now has kstrdup_const/kfree_const for reusing .rodata
(typically string literals) when possible; there's no reason to
duplicate that logic in the tracing system. Moreover, as the comment
above core_kernel_data states, it may not always return true for
.rodata - that is for example the case on x86_64, where we thus end up
kstrdup'ing all the passed-in strings.

Arguably, testing for .rodata explicitly (as kstrdup_const does) is
also more correct: I don't think one is supposed to be able to change
the name after creating the event_subsystem by passing the address of
a static char (but non-const) array.

Link: http://lkml.kernel.org/r/1441833841-12955-1-git-send-email-linux@rasmusvillemoes.dk
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
kernel/trace/trace_events.c